Goal seeking neural network architectures for image classification applications

Neural network approaches to image classification have become increasingly widespread in recent years, but have often been somewhat unsatisfactory for many practical applications because of the constraints imposed on processing speeds attainable, particularly in the implementation of appropriate training algorithms. Boolean neural networks are an alternative approach to neural network design. The Boolean neuron model has several advantages when compared to an analogue model: it is more easily implemented in hardware, uses faster learning algorithms and can realise non-linearly separable functions.<
>